### International Day of Philosophy - Prediction, Explanation, Emergence and Ontology

Today was the UNESCO's international day of philosophy.

I have deliberated for some time what I would post, and have now finally made up my mind.

The following will be a discussion of relation between predictability, determination, modelling and explanation, what role the concept of emergence plays and how this ought to affect our ontology, i.e. the set of beliefs pertaining to "what there is".

Recently, while reading through a thread on RDnet.com, my good friend Steve made the argument that we cannot predict, and thus not explain life arising from a universe of quarks - you cannot get from the fundamental laws of physics to specific predictions about populations, procreation, biodiversity etc. Thus, while we cannot predict and thus not explain it - we know it is certainly nothing "more than" physical "stuff" interacting. Or rather: We have sufficient reason to accept this as the hypothesis we go with for all intents and purposes, because it has the most explanatory value, because it is the most parsimonious and because there is no serious competition... "Magic" just isn't an explanation.

I would like to present my own thoughts on the subject - they differ in detail from Steve's opinion, but, I think, will clear up the matter by going into sufficient detail:

What we are dealing with are complex systems in the systems-theoretic term. While it is conceptually impossible for an single, isolated quark to have the property "temperature", and for a system with only one fundamental constituent to have entropy (the examples Steve used), it is not

Here, the findings and conceptual tools of complexity theory and information theory demonstrate my point: Kolmogorv Complexity, Bayesian prior probability, Shannon information entropy</i>and Minimum Description Length" stand in mathematical relation to each other - and we can meaningfully treat every real system as being fully described by the physical information it contains. As such, the entire history of that which treat as our universe up to now will be compeltely described by a listing the complete physical information of the system over time.

Now, readers who have some basic knowledge of physics will note that we have every reason to assume that ours is a probabilistic universe. This is certainly true. But it is not in conflict with my observations. We can certainly make out one unique past for this universe (even if others may "exist" in the quantum-multiverse).

Now, any one such finite history, since it is ex hypothesi describable (by describing the behaviour of all the elements and their intrinsic and relational properties through time) - it is also computable This means there is a shortest program that can run on some computer, either a Turing Machine or a Gödel Machine (see Schmidhuber's work on this here:) which will produce a complete description of the system in question.

Let's take the universe we observe - it has a definite, unique history. Given the initial conditions and all the rules that determine the system we

In any case, going with the multiverse-interpretation of Quantum-Mechanics we have a describable history of our universe - and a complete description can be computed.

It doesn't even matter that the hypothetical universal turing machine or Gödel-machine that calculates such a universe calculates the history of the universe on the most fundamental microlevel of physics. In fact, it is this property - the exhaustiveness - that guarantees that everything that happens in there is predictable. Thus, since our universe containing intelligent life is one universe with a definite, unique history - it is basically a computable universe.

The thing with aspects like life, populations, individual organisms and their behaviour is that they are not entities, variables that arise in the complete description: They are not things we quantify over in our complete description of the physical universe (since this will be on the level of strings in 11-dimensional space and time or fundamental particles).

Systems like life on earth, populations or individual organisms and facts like population-growth factors and they are not entities of physics - and thus not entities of a complete description of the physical universe. But they are included in this description nevertheless - as structures and patterns in the spatiotemporal history of the system, including the complete intrinsic and relational properties of everything that comprises the system in question (life, a population and its attributes or an organism and its history etc).

I have come up with what I think is a good analogy - the only difference between the situation I present and the "real world situation" is the complexity of the system, but what I show follows from the aspect of computabiltiy (in the technical sense), which the situation I present and the real world situation share:

Imagine a hypothetical universe whose complete description will take this mathematical form: "the set of complex c-values for which the orbit of 0 under iteration of the complex quadratic polynomial z

Now, is it possible to model, and thus to predict what the the geometrical, topological patterns will be that arise

The answer is: "Yes, but not without running a complete functional facsimile of this world". Or, to use a different phrase: "Yes, but not without actually plotting the function and then constituting that certain patterns/structures are there in these geometrical shapes".

Now, the determined, unique, history of our universe starting from the big bang and leading right up to you reading this sentence can be computed given every value for every variable that ever arises in this universe - a complete description. Given such a complete description - either in terms of initial conditions and rules for deterministic computation of future states, or by giving a complete extensional description of the system through time - this history is computable, and since there are only a finite number of intrinsic and relational properties, describable in finite descriptions - the machine computing it would only have to have finite capabilities (like memory and time to compute).

Through systems-theory, chaos theory etc we know how complex patterns and structures can form from quite simple initial conditions - in principle, the unique history of life on earth in our universe is something that we know can happen in a physical universe simply undergoing physical processes. This may sound unbelievable, but it follows logically. The only thing is - because life on earth is only an arbitrarily chosen part of a hugely complex, non-linear, dynamic system (though at the relevant level to predict life (the level of atoms and chemistry) it is reliable enough), you can only compute the precise history of that part of the system (taking the universe as one system and the history of life on earth as a part of it) when you have the

This means that while specifics about our universe and its inhabitants are necessarily practically impossible to derive or predict from what we know about the physical universe (because we haven't charted the complete spatiotemporal trajectory of every particle and wouldn't know how to extract the information we need), a physicalist reductionism, having no problem with this kind of weak emergence - is the most tenable position available.

This unpredictability is the case with systems that are like cellular automata - you can predict the future states of the system only by observing the system unfold - or by having a complete functional facsimile of it which you observe. Our universe with its period of rapid expansion, the incomplete annihilation of matter and antimatter, the symmetry and the breakings of symmetry is a highly complex, dynamic system. Over time, in such a system of fundamental particles, strings or whatever, structures will form... just letting the system with simple descriptions of how its parts behave run its course will in the end lead to such complex agglomerations of fundamental particles as starts, galaxies, planets and even living organisms.

The concept of self-organization does not only apply to such things as slime-mold behaviour, synchronized flashing of fireflies or fish schooling - we can also view the universe as a whole, consisting of fundamental particles (or strings - the actual preferred fundamental langugage of physics determines this) as a self-organizing system. After all - these fundamental constituents act and interact - or so science tells us - according to their physical propererties as we describe them. So all the properties e.g. a star has (including such high-level properties as the lifecycle-stage in which it is in) reduce to the interactions of fundamental constituents - they are emergent properties of the system of interacting fundamental particles over time - ascribed to the star as a whole, but completely reducible to the interactions of the fundamental particles involved.

Life on earth, consciousness and behaviour of individuals are the same - they emergent characteristics in this system because they are in fact high-level abstract sub-systems - patterns and structures that "are there" only in virtue of being embodied in behaviour of fundamental particles in the universe (or strings, or whatever).

As mentioned above, this is called "weak emergence", where it is accepted that there is no more to the system than that which is there when we describe it completely on the fundamental level - but the "things" we talk about are in fact high-level, interest relative structures and patterns in that system. (Having interest-relative boundaries is a hallmark of complex systems in systems theory).

The position of "non-reductive physicalism" advocates something called "strong emergence", which postulates that high-level properties do not reduce to the interactions of constituents of lower-levels of description of the system in question, yet are still properties of the system made up by the physical particles. (Consciousness, as such, would be a strongly emergent property for proponents of this view). This would mean some form of "magic", we would have the paradox of having to accept the fundamental level as that in which all is embodied - ie "all that is", and yet not as "all that is" because strongly emergent characteristics are not at all explicable/describable/predictable from the lower level.

My account above demonstrates how even the most complex and mind-boggling phenomena, like life on earth - can be understood as in principle describable, predictable and explicable from the fundamental level, but only when we are allowed to make reference to (and thus quantify over) structures and patterns in the descriptions of the system on the fundamental level - and high-level abstract properties.

Weak emergence is not opposed to reductive physicalism, eliminative materialism or functionalism. In fact, the notion that these are necessarily distinct approaches which propose opposing positions is simply a misunderstanding - a category-error if you will.

What we can do is gain explanatory insight into how our universe works through intertheoretic reduction and functional reduction. Scientific theories explaining our universe are always generalizing, and thus abstracting from the "real world" - but this is not a flaw... it is what makes explanation and understanding possible.

Our universe is a complex dynamic system so that no part of it can ever be perfectly, completely modelled and thus explained in every detail - without knowing everything about the system, every value of every variable. But by abstracting - by generalization, we can say that some observed event, like the ball on the pool table hitting another and driving it into the top-left pocket is explicable, describable through newtonian mechanics to sufficient detail. If we did not abstract and generalize, we would realize that the exact state of the bosons and leptons, of the matter-wave of the elements of this system are part of a complex, dynamic system where an immense number of values have to be known to describe it completely, far more than are possible to know and use for explanation and understanding.

And when we can - as e.g. John Bickle, Paul and Patricia Churchland and Daniel Dennett show - make actual explanatory inter-theoretic or functional reductions - we make explanatory progress - we enrich our understanding. This understanding and explaining necessarily involves abstraction, generalization. That's why we do not contradict ourselves by stating that certain interest-relative subsystems of entire system, as they are in all their detail (all intrinsic and relational properties), show emergent characteristics on high-levels of description.

Furthermore, eliminative materialism can be integrated as well, by recognizing that some conceptual structures we used and use to understand and explain the world around us are no longer viable, given the insight we have from the various sciences.

Thus we can integrate all these various approaches and provide a reasonable explanation of the relations between predictability, explicability, complex systems, dynamic and non-linear systems.

If my reasoning is sound, I will have lived up to at least one part of the praise UNESCO gives to philosophy, as its members have agreed for today (yesterday by the time I'm writing) to be the international day of philosophy: It can enlighten us, enhance our understanding of the universe, of the entire world around us and contribute substantially to the effort for peace and understanding. The latter is something a reasonable and humane ethical and political philosophy can do... the former is what I hope to be able to contribute to with this post.

All the best,

-Michael

I have deliberated for some time what I would post, and have now finally made up my mind.

The following will be a discussion of relation between predictability, determination, modelling and explanation, what role the concept of emergence plays and how this ought to affect our ontology, i.e. the set of beliefs pertaining to "what there is".

Recently, while reading through a thread on RDnet.com, my good friend Steve made the argument that we cannot predict, and thus not explain life arising from a universe of quarks - you cannot get from the fundamental laws of physics to specific predictions about populations, procreation, biodiversity etc. Thus, while we cannot predict and thus not explain it - we know it is certainly nothing "more than" physical "stuff" interacting. Or rather: We have sufficient reason to accept this as the hypothesis we go with for all intents and purposes, because it has the most explanatory value, because it is the most parsimonious and because there is no serious competition... "Magic" just isn't an explanation.

I would like to present my own thoughts on the subject - they differ in detail from Steve's opinion, but, I think, will clear up the matter by going into sufficient detail:

What we are dealing with are complex systems in the systems-theoretic term. While it is conceptually impossible for an single, isolated quark to have the property "temperature", and for a system with only one fundamental constituent to have entropy (the examples Steve used), it is not

*conceptually impossible*to derive the specifics of life on earth from a complete description of the universe in physical terms until now - in fact there has to be such a hypothethical complete description**so long as the history of our universe is computable**. Remember, we live in a finite universe with a finite past and finite constituents with finite relational and intrinsic properties. Given this, there exists a hypothetical complete description.Here, the findings and conceptual tools of complexity theory and information theory demonstrate my point: Kolmogorv Complexity, Bayesian prior probability, Shannon information entropy</i>and Minimum Description Length" stand in mathematical relation to each other - and we can meaningfully treat every real system as being fully described by the physical information it contains. As such, the entire history of that which treat as our universe up to now will be compeltely described by a listing the complete physical information of the system over time.

Now, readers who have some basic knowledge of physics will note that we have every reason to assume that ours is a probabilistic universe. This is certainly true. But it is not in conflict with my observations. We can certainly make out one unique past for this universe (even if others may "exist" in the quantum-multiverse).

Now, any one such finite history, since it is ex hypothesi describable (by describing the behaviour of all the elements and their intrinsic and relational properties through time) - it is also computable This means there is a shortest program that can run on some computer, either a Turing Machine or a Gödel Machine (see Schmidhuber's work on this here:) which will produce a complete description of the system in question.

Let's take the universe we observe - it has a definite, unique history. Given the initial conditions and all the rules that determine the system we

**can**predict any later stage - this was LaPlace's insight. Remember, the multiverse interpretation of QM does not postulate true randomness, it is deterministic - qm-phenomena in our universe appear random (indeterministic) because we cannot know beforehand which universe-history in the multiverse we inhabit with respect to the outcome of the qm-uncertain event in question.In any case, going with the multiverse-interpretation of Quantum-Mechanics we have a describable history of our universe - and a complete description can be computed.

It doesn't even matter that the hypothetical universal turing machine or Gödel-machine that calculates such a universe calculates the history of the universe on the most fundamental microlevel of physics. In fact, it is this property - the exhaustiveness - that guarantees that everything that happens in there is predictable. Thus, since our universe containing intelligent life is one universe with a definite, unique history - it is basically a computable universe.

The thing with aspects like life, populations, individual organisms and their behaviour is that they are not entities, variables that arise in the complete description: They are not things we quantify over in our complete description of the physical universe (since this will be on the level of strings in 11-dimensional space and time or fundamental particles).

Systems like life on earth, populations or individual organisms and facts like population-growth factors and they are not entities of physics - and thus not entities of a complete description of the physical universe. But they are included in this description nevertheless - as structures and patterns in the spatiotemporal history of the system, including the complete intrinsic and relational properties of everything that comprises the system in question (life, a population and its attributes or an organism and its history etc).

I have come up with what I think is a good analogy - the only difference between the situation I present and the "real world situation" is the complexity of the system, but what I show follows from the aspect of computabiltiy (in the technical sense), which the situation I present and the real world situation share:

Imagine a hypothetical universe whose complete description will take this mathematical form: "the set of complex c-values for which the orbit of 0 under iteration of the complex quadratic polynomial z

_{n+1}= z_{n}^{2}+ c remains bounded"Now, is it possible to model, and thus to predict what the the geometrical, topological patterns will be that arise

*when you plot this to get the Mandelbrot-set we all know?*The answer is: "Yes, but not without running a complete functional facsimile of this world". Or, to use a different phrase: "Yes, but not without actually plotting the function and then constituting that certain patterns/structures are there in these geometrical shapes".

Now, the determined, unique, history of our universe starting from the big bang and leading right up to you reading this sentence can be computed given every value for every variable that ever arises in this universe - a complete description. Given such a complete description - either in terms of initial conditions and rules for deterministic computation of future states, or by giving a complete extensional description of the system through time - this history is computable, and since there are only a finite number of intrinsic and relational properties, describable in finite descriptions - the machine computing it would only have to have finite capabilities (like memory and time to compute).

**Because we know**that every pattern in any complex system whose history is computable is determined and determinable from the lowest level of description - we know that the patterns and structures we call "life"**are explained**by fundamental physics!Through systems-theory, chaos theory etc we know how complex patterns and structures can form from quite simple initial conditions - in principle, the unique history of life on earth in our universe is something that we know can happen in a physical universe simply undergoing physical processes. This may sound unbelievable, but it follows logically. The only thing is - because life on earth is only an arbitrarily chosen part of a hugely complex, non-linear, dynamic system (though at the relevant level to predict life (the level of atoms and chemistry) it is reliable enough), you can only compute the precise history of that part of the system (taking the universe as one system and the history of life on earth as a part of it) when you have the

**total information**about the entire system, which in our case would be the entire universe. Now, the totality of this information is there, but as a part of this system, we can never get at the whole of the information.This means that while specifics about our universe and its inhabitants are necessarily practically impossible to derive or predict from what we know about the physical universe (because we haven't charted the complete spatiotemporal trajectory of every particle and wouldn't know how to extract the information we need), a physicalist reductionism, having no problem with this kind of weak emergence - is the most tenable position available.

This unpredictability is the case with systems that are like cellular automata - you can predict the future states of the system only by observing the system unfold - or by having a complete functional facsimile of it which you observe. Our universe with its period of rapid expansion, the incomplete annihilation of matter and antimatter, the symmetry and the breakings of symmetry is a highly complex, dynamic system. Over time, in such a system of fundamental particles, strings or whatever, structures will form... just letting the system with simple descriptions of how its parts behave run its course will in the end lead to such complex agglomerations of fundamental particles as starts, galaxies, planets and even living organisms.

The concept of self-organization does not only apply to such things as slime-mold behaviour, synchronized flashing of fireflies or fish schooling - we can also view the universe as a whole, consisting of fundamental particles (or strings - the actual preferred fundamental langugage of physics determines this) as a self-organizing system. After all - these fundamental constituents act and interact - or so science tells us - according to their physical propererties as we describe them. So all the properties e.g. a star has (including such high-level properties as the lifecycle-stage in which it is in) reduce to the interactions of fundamental constituents - they are emergent properties of the system of interacting fundamental particles over time - ascribed to the star as a whole, but completely reducible to the interactions of the fundamental particles involved.

Life on earth, consciousness and behaviour of individuals are the same - they emergent characteristics in this system because they are in fact high-level abstract sub-systems - patterns and structures that "are there" only in virtue of being embodied in behaviour of fundamental particles in the universe (or strings, or whatever).

As mentioned above, this is called "weak emergence", where it is accepted that there is no more to the system than that which is there when we describe it completely on the fundamental level - but the "things" we talk about are in fact high-level, interest relative structures and patterns in that system. (Having interest-relative boundaries is a hallmark of complex systems in systems theory).

The position of "non-reductive physicalism" advocates something called "strong emergence", which postulates that high-level properties do not reduce to the interactions of constituents of lower-levels of description of the system in question, yet are still properties of the system made up by the physical particles. (Consciousness, as such, would be a strongly emergent property for proponents of this view). This would mean some form of "magic", we would have the paradox of having to accept the fundamental level as that in which all is embodied - ie "all that is", and yet not as "all that is" because strongly emergent characteristics are not at all explicable/describable/predictable from the lower level.

My account above demonstrates how even the most complex and mind-boggling phenomena, like life on earth - can be understood as in principle describable, predictable and explicable from the fundamental level, but only when we are allowed to make reference to (and thus quantify over) structures and patterns in the descriptions of the system on the fundamental level - and high-level abstract properties.

Weak emergence is not opposed to reductive physicalism, eliminative materialism or functionalism. In fact, the notion that these are necessarily distinct approaches which propose opposing positions is simply a misunderstanding - a category-error if you will.

What we can do is gain explanatory insight into how our universe works through intertheoretic reduction and functional reduction. Scientific theories explaining our universe are always generalizing, and thus abstracting from the "real world" - but this is not a flaw... it is what makes explanation and understanding possible.

Our universe is a complex dynamic system so that no part of it can ever be perfectly, completely modelled and thus explained in every detail - without knowing everything about the system, every value of every variable. But by abstracting - by generalization, we can say that some observed event, like the ball on the pool table hitting another and driving it into the top-left pocket is explicable, describable through newtonian mechanics to sufficient detail. If we did not abstract and generalize, we would realize that the exact state of the bosons and leptons, of the matter-wave of the elements of this system are part of a complex, dynamic system where an immense number of values have to be known to describe it completely, far more than are possible to know and use for explanation and understanding.

And when we can - as e.g. John Bickle, Paul and Patricia Churchland and Daniel Dennett show - make actual explanatory inter-theoretic or functional reductions - we make explanatory progress - we enrich our understanding. This understanding and explaining necessarily involves abstraction, generalization. That's why we do not contradict ourselves by stating that certain interest-relative subsystems of entire system, as they are in all their detail (all intrinsic and relational properties), show emergent characteristics on high-levels of description.

Furthermore, eliminative materialism can be integrated as well, by recognizing that some conceptual structures we used and use to understand and explain the world around us are no longer viable, given the insight we have from the various sciences.

Thus we can integrate all these various approaches and provide a reasonable explanation of the relations between predictability, explicability, complex systems, dynamic and non-linear systems.

If my reasoning is sound, I will have lived up to at least one part of the praise UNESCO gives to philosophy, as its members have agreed for today (yesterday by the time I'm writing) to be the international day of philosophy: It can enlighten us, enhance our understanding of the universe, of the entire world around us and contribute substantially to the effort for peace and understanding. The latter is something a reasonable and humane ethical and political philosophy can do... the former is what I hope to be able to contribute to with this post.

All the best,

-Michael

(Anonymous)You're a smart manBut well done anyway. I'm glad there's smart buggers out there like you.

Brian.